Unvoiced Plosives Research Articles

Respiratory laryngeal coordination during vowel-plosive-vowel transition in shouted speech

This study investigates differences in respiratory-laryngeal coordination between normal and shouted speech in the production of a vowel-plosive-vowel utterance, with the goal toward synthesis of such utterances in loud speech. The results showed that compared to normal condition, shouted speech produced higher intraoral pressure during vocal tract closure, higher airflow at the release of vocal tract closure, increased fundamental frequency and increased duration of glottal closure during the following vowel, and increased respiratory activities. For unvoiced plosives, shouted speech also produced increased voice onset time of the following vowel, but its importance to the perception of shouted speech was small. Computational simulations further showed that such increased voice onset time is likely due to increased maximum vocal fold abduction and/or delayed initiation of vocal fold adduction, which are necessary to avoid voice onset immediately after the release of vocal tract closure in conditions of high subglottal pressure.

Swiss German dialects spoken by second-generation immigrants: bilingual speech and dialect transformation

ABSTRACTGerman-speaking Switzerland is characterised by a type of diglossia where the ‘low variety’ is used in almost all domains of everyday communication. For children of immigrant workers, a Swiss German dialect normally constitutes one of their two first languages. Nevertheless, one can observe two different scenarios how the dialect enters the repertoire of these multilingual subjects.The first scenario entails the combined use of a heritage language (e.g. Italian) and a Swiss German dialect. Code-switching among peers of the same ethnic group serves a wide array of discourse and social functions, above all the expression of a composite bicultural identity, but the employed varieties do not differ structurally from those of monolingual speakers.The second scenario, instead, gives rise to a structural transformation of the traditional dialects. For instance, an individual speaker may be perceived as originating from a particular region of Switzerland on the basis of his or her vowel qualities, whereas at the same time the voicing of (normally unvoiced) plosives makes him or her recognisable as a person with an immigrant background. Our analysis focuses on these linguistic transformations of dialects into ‘(multi-)ethnolectal’ dialects, whose social function seem to lie in the signalling of an allochthonous (‘non-Swiss’) identity.

6 トランスパラタルアーチ装着により影響をうける無声破裂音の経日的変化(第481回 大阪歯科学会例会)

6 トランスパラタルアーチ装着により影響をうける無声破裂音の経日的変化(第481回 大阪歯科学会例会)

A scale–rate filter selection method in the spectro-temporal domain for phoneme classification

Recently, there has been a significant increase in studies employing auditory models in speech recognition systems. In this paper, we propose a new evolutionary tuned feature extraction method by spectro-temporal analysis. In our proposed model, there is a special subspace for each phoneme with a specific best scale in the spectral filter and a specific best rate in the temporal filter. These two parameters were obtained by genetic cellular automata evolutionary algorithm. The extracted features from the specific subspace are classified by a binary one-versus-rest support vector machine. Finally, a multiclass classifier for all phonemes is employed by combining these sub-models. The proposed method improved the discrimination of phonemes significantly especially in highly confusable phonemes. To show the efficiency of the proposed feature sets, it was empirically compared with two baseline models. The achieved relative improvements are about 10% in classification rate for voiced plosives, unvoiced plosives and nasals; and about 7.38% for front vowels relative to the state of the art baseline model.

C012 The results of combined electric acoustic stimulation in children with ski slope deafness using soft surgery techniques to preserve low frequency hearing

Recently, there has been a significant increase in studies employing auditory models in speech recognition systems. In this paper, we propose a new evolutionary tuned feature extraction method by spectro-temporal analysis. In our proposed model, there is a special subspace for each phoneme with a specific best scale in the spectral filter and a specific best rate in the temporal filter. These two parameters were obtained by genetic cellular automata evolutionary algorithm. The extracted features from the specific subspace are classified by a binary one-versus-rest support vector machine. Finally, a multiclass classifier for all phonemes is employed by combining these sub-models. The proposed method improved the discrimination of phonemes significantly especially in highly confusable phonemes. To show the efficiency of the proposed feature sets, it was empirically compared with two baseline models. The achieved relative improvements are about 10% in classification rate for voiced plosives, unvoiced plosives and nasals; and about 7.38% for front vowels relative to the state of the art baseline model.

Back-and-Forth Methodology for Objective Voice Quality Assessment: From/to Expert Knowledge to/from Automatic Classification of Dysphonia

This paper addresses voice disorder assessment. It proposes an original back-and-forth methodology involving an automatic classification system as well as knowledge of the human experts (machine learning experts, phoneticians, and pathologists). The goal of this methodology is to bring a better understanding of acoustic phenomena related to dysphonia. The automatic system was validated on a dysphonic corpus (80 female voices), rated according to the GRBAS perceptual scale by an expert jury. Firstly, focused on the frequency domain, the classification system showed the interest of 0–3000 Hz frequency band for the classification task based on the GRBAS scale. Later, an automatic phonemic analysis underlined the significance of consonants and more surprisingly of unvoiced consonants for the same classification task. Submitted to the human experts, these observations led to a manual analysis of unvoiced plosives, which highlighted a lengthening of VOT according to the dysphonia severity validated by a preliminary statistical analysis.

Aussprachestörungen bei Kindern mit spezifischen Sprachentwicklungsstörungen und schwerhörigen Kindern im Vergleich

Children with hearing impairment commonly present with articulation deficits. This is considered one main characteristic. 24 pairs of children (5;00 - 6;11 yrs) with articulation deficits in the context of a language disorder were matched in the severity of the language disorder, age, gender and nonverbal intelligence. Of each matched pair one child had a hearing impairment, the other a specific language impairment. Both groups were compared for their articulation. In children with hearing impairment the influence of the severity of the hearing loss was evaluated. In both groups the same type of articulation deficit for consonants was found. Both groups of children had most frequently problems with sibilants. Children with specific language impairment had slightly more severe articulation deficits. In unvoiced plosives children with specific language impairment demonstrated more often substitutions. The severity of the language disorder, but not the severity of the articulation deficit, was correlated with the degree of hearing loss. Articulation deficits in children with hearing impairment are slightly less than in children with specific language impairment, but follow the same pattern.

A produção de plosivas por crianças de três anos falantes do português brasileiro

OBJETIVO: apresentar as características fonético-acústicas de produção das consoantes plosivas vozeadas e não-vozeadas por crianças de três anos. MÉTODOS: análise qualitativa baseada na inspeção das características fonético-acústicas de eventos fônicos ocorridos no intervalo de obstrução, na liberação da plosão e na transição para a vogal e análise quantitativa das medidas do Voice Onset Time (VOT), de três repetições das seis consoantes plosivas do Português Brasileiro (PB), por quatro crianças, sendo três meninos e uma menina, falantes nativos do PB, por meio de análise fonético-acústica, em espectrograma de banda larga. Os achados foram comparados com estimativas de valores para a faixa adulta de falante nativo do PB. RESULTADOS: as crianças apresentaram produções semelhantes às que são encontradas na fala adulta, tanto para as plosivas vozeadas como para as não-vozeadas, bilabiais, alveolares e velares. Apresentaram também plosivas vozeadas com interrupção de barra de sonoridade e ocorrência de VOT positivo, e qualidade diferenciada de burst. Para as não-vozeadas, ocorreram a presença da breathy vowel no período de obstrução das plosivas e a presença de aspiração no segmento correspondente ao VOT positivo, para os três pontos de articulação pesquisados. CONCLUSÃO: as diferenças encontradas na produção das plosivas vozeadas, quanto às características da barra de sonoridade e qualidade do burst, a presença da breathy vowel nas plosivas não-vozeadas e aspiração, caracterizam a produção das plosivas nessa faixa etária. A diversidade das medidas acústicas, verificada pelos valores dos desvios-padrão sugerem que as crianças realizaram diversas tentativas, com diferentes ajustes para produzir tais sons.

Acoustic Analysis of Consonants in Whispered Speech

An acoustic analysis of whispered consonants in comparison to normally phonated consonants was conducted in time and intensity domains. Consonant duration and average root mean square intensity were measured for six speakers in both articulation modes. Each of 25 Serbian consonants (C) was sited between the vowel /a/ forming a syllable of /aCa/ type. Such a syllable was placed in initial, medial, and final position in the carrier sentence. Results showed that whispered consonants have a prolonged duration of about 10% on average (statistically significant, ANOVA test), and that the unvoiced consonants have a smaller time dimension extension (5.8%) than voiced ones (15.3%). Examination at subphonemic level showed that there is no difference in voice-onset-time and affrication duration in unvoiced plosives and affricates, in both whispered and phonated mode of articulation, but the difference is significant for voiced ones. Analysis of consonant duration versus place of articulation showed that palatal place is most sensitive in the process of whispering. In all experiments, the results are very consistent with respect to the subjects and test material (Pearson's correlation was between 0.6 and 0.9). In intensity domain, all unvoiced consonants in whispered mode of articulation have almost unchanged intensity in comparison to phonated mode (the difference is maximum 3.5 dB). On the contrary, voiced consonants in the whispered mode were reduced in intensity by as much as 25 dB, as nasals and semivowels. Average intensity of whispered consonants is lowered by 12d B in comparison to phonated ones, and does not depend on syllabic position inside the sentences.

Phonology impacts segmentation in online speech processing

Peña, Bonatti, Nespor, and Mehler (2002) investigated an artificial language where the structure of words was determined by nonadjacent dependencies between syllables. They found that segmentation of continuous speech could proceed on the basis of these dependencies. However, Peña et al.’s artificial language contained a confound in terms of phonology, in that the dependent syllables began with plosives and the intervening syllables began with continuants. We consider three hypotheses concerning the role of phonology in speech segmentation in this task: (1) participants may recruit probabilistic phonotactic information from their native language to the artificial language learning task; (2) phonetic properties of the stimuli, such as the gaps that precede unvoiced plosives, can influences segmentation; and (3) grouping by phonological similarity between dependent syllables contributes to learning the dependency. In a series of experiments controlling the phonological and statistical structure of the language, we found that segmentation performance is influenced by the three factors in different degrees. Learning of nonadjacent dependencies did not occur when (3) is eliminated. We suggest that phonological processing provides a fundamental contribution to distributional analysis.

Changes over time in unvoiced plosives after placement of transpalatal arch

Changes over time in unvoiced plosives after placement of transpalatal arch

Aging of hearing ability

The average life span of the Japanese has been drastically extended in recent decades thanks to developments in medical science and treatments. Considerations for the aged are demanded in every aspect of social life including voice announcements by synthetic speech. The real situation of hearing synthetic speech by aged persons is clarified through dictation tests on synthetic speech generated by a commercial synthesizer system using texts from Murphy’s law for 58 subjects aged 10 to 80 years old. In order to investigate the effects of hearing loss and language ability, dictation results were compared with results of hearing tests for sinusoidal waves and language ability tests by various types of written and oral tests. Dictation scores of primary school children and aged persons show larger variances compared with those for subjects 20–59 years old. Rates of mishearing by aged persons having hearing difficulty sometimes exceed 50% in mora units, although those of primary school children remain within 18% at the most. The correlation coefficient between dictation score and comprehension score was 0.76. Occurrence rates of phonemic confusion on plosive sounds by subjects with hearing difficulty show clear differences between voiced plosives and unvoiced plosives. [Work supported by STA, Japan.]

Occlusive Silence Duration of Voiceless Intervocalic Plosives and Voicing Perception by Normal and Hearing-Impaired Subjects

The duration of occlusive silence of intervocalic unvoiced plosives is one of several acoustic cues contributing to the correct identification of voicelessness for normally hearing subjects. In a previous investigation on one velar plosive, some hearing-impaired subjects required an abnormally long duration of its occlusive silence in order to avoid voicing confusions. The present study investigated whether this would occur for the different plosives in various intervocalic contexts and for male and female voices. Speech signals from a male and a female voice pronouncing the three voiceless plosives (p,t,k) in three different vocalic contexts (a,i,u) were recorded. Each of these 18 vowel-consonant-vowel (VCV) utterances was edited to obtain a series of tokens with occlusive silence durations varying from 0 to 180 msec. Phonetic identification of these tokens was assessed for seven normally hearing persons and for seven subjects with a sensorineural hearing loss. For the various VCVs, about half of the hearing impaired subjects required abnormally long occlusive silence durations to avoid voicing confusions. For normal as well as hearing-impaired subjects duration thresholds varied considerably and not systematically for the different VCVs. The amount of voicing confusions tended to be related to audiogram loss. For hearing-impaired subjects, a normal occlusive silence duration can lead to voicing misperceptions for the various plosives in different vowel contexts and for male and female voices. The large variability in the results of each subject for the different VCVs indicates that even a strong anomaly in temporal perception, which certainly underlies a need for abnormally long occlusive silence duration, did not systematically overrule voicing perception.

Phonetic classification of plosive voicing using computational modeling

One approach to investigating the roles of auditory and phonetic processing in speech perception tasks is to compare the performance of listeners with no phonetic processing capabilities (animals and machines) to those of human listeners. Here, a computational model of the peripheral auditory system was coupled to an artificial neural net pattern classifier in order to investigate the essential auditory and phonetic features that distinguish voiced and unvoiced plosives in initial position. The computational model and the human listeners were presented with three stimulus series, varying in voice onset time, representing English bilabials (‘‘ba’’–‘‘pa’’), alveolars (‘‘da’’–‘‘ta’’), and velars (‘‘ga’’–‘‘ka’’). The categorization functions obtained for the model and for human listeners will be compared with regard to their sharpness, and shift of phoneme boundary with place of articulation. [Work supported by the Joint Council Initiative in Cognitive Science/HCI (UK).]

Speech visualization and its application for the hearing impaired

A speech‐visualization system has been developed as an extension of research on a color display system for connected speech [Watanabe et al., IEEE Trans. Acoust. Speech Signal Process. ASSP‐33, 164–173 (1985)]. In this newly developed system, normalized spectrographic patterns are overlapped as a change of luminance on a whole pattern that represents voiced segments by colors. The spectrographic patterns are represented with high luminance in unvoiced portions, and with low luminance in voiced ones so as not to disturb the color image. This extension allows unvoiced plosives as well as fricatives and affricates to be clearly visualized by the sharp contrast between their black‐and‐white patterns and the vivid colors of the voiced components. Moreover, in order to generate a visually clear image, a cleaner, which eliminates unnecessary patterns such as aspirates that are often observed at the end of utterances, is introduced. Finally, two applications of this system are described. One is a speech training system in which a personal computer, with a database, controls the speech‐visualization system and provides a hearing aid. The other is a telecommunication system that utilizes a public telephone channel. [This work was supported by TAF.]

Speaker‐independent recognition of unvoiced plosives using a convex time pattern of a posteriori probability

This paper describes a method of speaker‐independent recognition of unvoiced plosives. In this method, the convex characteristics of the time pattern of a posteriori probability is adopted for eliminating effects of speaker difference and coarticulation. The time pattern of a posteriori probability is more suitable than that of the distance pattern. In the first stage of the method, a posteriori probabilities for four categories (/p/, /t/, /k/, and silence) are calculated frame by frame from a five‐channel spectrum of five time frames using the Bayes theorem. In the next stage, the convex part of the time pattern of a posteriori probability is decided as an unvoiced plosive. The decision using the dynamic characteristics is more suitable for speaker‐independent recognition than that using static threshold. The recognition experiments are conducted for about 1400 samples of unvoiced plosives in 166 Japanese city words uttered by five male speakers. These experiments are carried out under the condition of automatic phoneme detection and without the knowledge of the following vowel. The recognition rate of 81% is obtained for the speaker‐dependent case and 59% for the speaker‐independent case. [Work supported by Grant‐in‐Aid for Scientific Research on Priority Areas, The Ministry of Education, Science and Culture of Japan.]

The effect of wearing aerodynamic measuring devices on speech articulation

Aerodynamic measures of speech production have been used to obtain insights into the dynamic properties of the vocal tract during speech production. It is unclear, however, how wearing devices such as the pneumotachograph and oral pressure transducers affects the articulatory processes being studied. The present research examined the articulatory characteristic of speech production by two normal speakers under two conditions: (1) while wearing a Rothenberg type pneumotachograph mask and oral pressure transducer and (2) without these devices. The x-ray microbeam data were obtained for lip, jaw, and tongue movement during multiple productions of voiced and unvoiced plosives in varied vowel contexts. Preliminary analysis suggests that articulatory dynamics appear not to change as a result of wearing the mask, although movement amplitudes may. Comparisons between the two conditions will be presented, together with implications for use of such measures as indices of speech articulatory behavior.

Duration as a function of sound pressure level

The association between sound pressure level and duration of French vowels and consonants was calculated using the Pearson products moment correlation. The association between sound pressure and duration was highly significant and positive for vowels, equally significant but negative for unvoiced plosives and fricatives; negative and somewhat less significant for voiced plosives and fricatives; non-significant for liquids. The regression lines for phonemes belonging to the same class are very similar. All vowels are characterized by a steeply ascending regression line; consonants, excepted liquids, by a decending regression line. The results are interpreted in terms of a chronemic strategy with the aim of conveying the message in spite of high background noise.

Coarticulation and the identification of initial and final plosives

Natural CVt and tVC utterances (V = |i,u,α|; C = |p,t,k,b,d,g|) of one American English male speaker were used to evaluate which information in CV and VC syllables is perceptually relevant for the identification of initial and final plosives. Different initial or final parts of the utterances were deleted and replaced by silence or by a 300-ms burst of noise. Subjects were asked to identify the (deleted) plosives. The noise paradigm was introduced to make the incomplete (C)Vt or tV(C) syllables sound more natural: it improved the correct score for initial plosives. Identification results are interpreted in articulatory and spectral terms. For initial unvoiced plosives most coarticulation takes place in the aspiration part and hardly any information about the preceding plosive is preserved in the vowel (transition) part. For the heavily truncated (C)ut stimuli, subjects give mainly |b| responses, while |p| is the favorable response for all heavily truncated tV(C) stimuli. [Research carried out at Speech Communications Research Laboratory, Inc., Santa Barbara and supported by the Netherlands Organization for the Advancement of Pure Research (ZWO).]